Whitespace in text tokenized as IGNORABLE_WHITESPACE in XmlReader

[westnordost]

When I have this XML

<user>dude &amp; &lt;dudette&gt;</user>

I expect to get the following events when I iterate through it via the XmlReader:

1. START_DOCUMENT
2. START_ELEMENT localName="user"
3. TEXT text="dude "
4. ENTITY_REF text="&"
5. TEXT text=" "
6. ENTITY_REF text="<"
7. TEXT text="dudette"
8. ENTITY_REF text=">"
9. END_ELEMENT localName="user"
10. END_DOCUMENT

However, number 5 doesn't turn up as a TEXT but as an IGNORABLE_WHITESPACE.

I think this is a bug, this is not an ignorable whitespace. Whitespaces between XML elements, such as <user>abc</user> <id>1234</id> would be ignorable.

(By the way, the existence of CDSECT and ENTITY_REF was a pitfall (aka footgun) for me, I assumed before that the XMLReader would already have all text content, i.e. I expected there would be just TEXT text="dude & <dudette>" and then END_ELEMENT.)

[westnordost]

Note on the "by the way":

Just tested the behavior of org.xmlpull.v1.XmlPullParser:

When XmlPullParser.getEventType() is XmlPullParser.TEXT, XmlPullParser.getText() indeed returns the entire string, i.e. in the above example "dude & <dudette>", despite event types for CDSECT and ENTITY_REF also existing... 🤔
(The relevant documentation reveals that XmlPullParser has somewhat of two APIs for iterating through the events, one a bit more low level than the other)

[pdvrieze]

Sorry about the delay. I've just fixed it in dev. This will still parse it at separate events, but will note that the whitespace is delimited by entities and thus not (attempt to) detect whitespace (and thus not generate ignorable whitespace events).